Scaling a variable by another variable

[Andreas Andereggen]

Hi everyone,

I need to calculate a variable as the standard deviation of another variable scaled by yet another variable, requiring at least 3 years of data.
This is how I calculate the standard deviation:

egen cfv = sd(oibdp)

But then I don't know how to proceed to scale the variable as explained above. Can anyone help me?

Thank you in advance!

[Nick Cox]

What you want isn't clear to me.

egen cfv = sd(oibdp)

uses all available data for one variable and calculates a single standard deviation which goes in a new variable, each observation holding the same value.

Nothing there is said about years of available data.

I think you need to tell us much more about your data. https://www.statalist.org/forums/help#stata is key.

[Andreas Andereggen]

Sorry about my unclear post.

I am using the Compustat data set, where I need to calculate the Cash flow volatility. It is said to be the standard deviation of historical operating income, scaled by total assets, requiring at least 3 years of data.
Here an example of what my data looks like:

gvkey fyear at oibdp
1000 1970 33.45 5.386
1000 1971 29.33 2.512
1000 1972 19.907 4.109
1000 1973 21.771 4.514
1000 1974 25.638 4.82
1000 1975 23.905 6.364
1000 1976 38.586 8.741
1000 1977 44.025 6.01
1001 1983 14.08 2.65
1001 1984 16.267 3.208
1001 1985 39.495 7.247
1002 1970 13.722 1.84
1002 1971 12.123 2.009
1002 1972 15.676 1.968
1003 1983 8.529 2.138
1003 1984 8.241 .825
1003 1986 14.586 2.462
1003 1987 16.042 .111
1003 1988 16.28 -3.68
1003 1989 10.109 -1.532
1004 1967 4.88 .201
1004 1968 10.402 .999
1004 1969 11.584 1.149
1004 1970 13.468 1.574
1004 1971 16.501 2.207
1004 1972 24.068 3.376
1004 1973 31.584 5.075
1004 1974 43.353 4.785
1004 1975 43.539 4.958
1004 1976 45.55 5.538
1004 1977 56.502 6.74
1004 1978 64.194 10.037
1004 1979 73.758 10.556
1004 1980 83.075 11.058
1004 1988 356.391 51.051
1004 1989 388.521 54.032

with gvkey : firm id, fyear: time id, at: total assets, oidp: operating income before depreciation.

Is it more clear like this?

[Nick Cox]

I guess that helps someone. I don't work in your field but I imagine that lumping together different gvkey makes no sense. Beyond that the questions you need to answer for others to help you might well include

sd(operating income)/assets OR sd(operating income/assets)?

in 3 year moving windows?

[Phil Bromiley]

I suggest following the FAQ on asking questions. For example, using dataex to generate your data will help us use your data.

I think I understand what you are trying to do. You want the 3 year standard deviation in operating income by firm year. I assume this is the current and two previous years.

xtset  gvkey fyear

g Lat=L.at
g L2at=L2.at

egen std1=rowsd(at Lat L2at) if Lat<. & L2at<.
g byat=std1/at

I suspect there is a more efficient way to do this with rangestat or something, but this works. You have to choose whether it is current and previous two years or previous 3 years and what to do for the second observation for each firm (where you only have 2 years of data - I've set it to missing if you if you don't have both lagged variables.

As Nick notes, there is also the possibility that you want to normalize income before doing the std, but that is so easy to program, you should be able to do it.